Summary
Reporting a ChatGPT + DevSpace integration/model-selection compatibility problem. The intended model is 6 Pro, but explicitly choosing GPT-5.6 Sol and then Pro selects 5.6 Pro, not 6 Pro.
This is not yet established as a DevSpace server defect: ChatGPT owns the model picker, and Oracle browser automation is involved. Please help clarify the supported 6 Pro workflow and document it for DevSpace users.
Environment
- Windows; ChatGPT web with a registered DevSpace MCP app
@waishnav/devspace 1.0.8 (npm latest checked 2026-09-07)@steipete/oracle 0.18.0 (npm latest checked 2026-09-07)- Our local integration has compatibility patches; this is not a clean upstream-only reproduction.
Observed model-selection behavior
- Open ChatGPT's composer model/effort picker.
- Explicitly select Latest (
최신 in Korean), then choose Pro at maximum power (5/5).
- The composer displays 6 Pro. Before selecting Pro, Latest may have no numeric version label.
- If GPT-5.6 Sol is explicitly selected instead, choosing Pro displays 5.6 Pro.
Integration guidance or automation that pins the 5.6 row therefore does not select the intended 6 Pro model. Merely preserving the existing selection is also insufficient when 5.6 was selected previously: the flow must explicitly choose Latest before Pro.
Expected behavior / requested guidance
- Document and support explicit Latest selection → Pro → verify the visible 6 Pro label for DevSpace-connected ChatGPT sessions.
- Do not conflate Latest with the explicit GPT-5.6 Sol row.
- Clarify whether any DevSpace setup/tool-exposure changes are required for 6 Pro, or whether this belongs entirely to ChatGPT/Oracle.
- If upstream-owned, document the workaround or cross-reference the upstream issue so model-selection failures are not misdiagnosed as MCP failures.
Workaround and scope
We are implementing explicit Latest selection followed by Pro, with fail-closed verification of model and effort. We intend to retain that route until Oracle officially supports 6 Pro and the native route is actually verified.
This report concerns selecting the intended model. It does not claim that DevSpace can never work with 6 Pro, nor that selecting Pro proves actual workspace capability. No credentials, private workspace paths, or conversation URLs are included.
Summary
Reporting a ChatGPT + DevSpace integration/model-selection compatibility problem. The intended model is 6 Pro, but explicitly choosing GPT-5.6 Sol and then Pro selects 5.6 Pro, not 6 Pro.
This is not yet established as a DevSpace server defect: ChatGPT owns the model picker, and Oracle browser automation is involved. Please help clarify the supported 6 Pro workflow and document it for DevSpace users.
Environment
@waishnav/devspace1.0.8 (npm latest checked 2026-09-07)@steipete/oracle0.18.0 (npm latest checked 2026-09-07)Observed model-selection behavior
최신in Korean), then choose Pro at maximum power (5/5).Integration guidance or automation that pins the 5.6 row therefore does not select the intended 6 Pro model. Merely preserving the existing selection is also insufficient when 5.6 was selected previously: the flow must explicitly choose Latest before Pro.
Expected behavior / requested guidance
Workaround and scope
We are implementing explicit Latest selection followed by Pro, with fail-closed verification of model and effort. We intend to retain that route until Oracle officially supports 6 Pro and the native route is actually verified.
This report concerns selecting the intended model. It does not claim that DevSpace can never work with 6 Pro, nor that selecting Pro proves actual workspace capability. No credentials, private workspace paths, or conversation URLs are included.